The Grammar of GraphicsThe first speaker, Conrad Taylor, introduced us to the Grammar of Graphics. He opened with a Venn Diagram showing the intersection of Information Designers, IT people, and Information and Knowledge Managers. He pointed out that they don’t spend enough time talking to each other and the whole information community would benefit from fewer silos and more inter-disciplinary engagement.
Visual reasoning is aeons old, and common to most of the animal kingdom. Humans navigate the world and gain huge amounts of information in a “blink of the eye”. However, there has been little research on the grammar and semantics of presenting visual information. Communication goes beyond use of language, but includes visual cues such as gesturing; drawing is often an extended form of gesturing, with the movement conveying the meaning perhaps more so that the resulting sketches. However, there has been a lack of research into visual intelligence grammar of images and semantics of imagery.
Often the ownership of the technology affects the nature of the communication, so tools like Mind Maps are a commercial product. However, in pre-literate communities creating graphics by drawing in the sand and using stones to represent ideas can be very effective and egalitarian as no-one owns the technology of expression.
Conrad offered an overview of the history of information visualization, mentioning the Tree of Porphyry, medical diagrams by Ibn al-Nafis, Agricola’s De Re Metallica, Ebstorfer’s Mappa Mundi, and Mohammad Al-Idrissi’s map of Europe. In the 18th and 19th centuries people started to produce data graphics, examples include Joseph Priestley’s biographical timelines, William Playfair’s line, bar, and pie charts, Charles Joseph Minard’s graphic (1869) showing the advance and retreat of Napoleon’s army, John Snow’s map of cholera cases in London, Otto and Marie Neurath’s diagrams, and Florence Nightingale’s coxcomb charts – an early example of “socioinformatics”.
There are pitfalls with graphics, however, choropleth maps can mislead and designers will often think they are scaling up linearly – e.g. to make a shape twice as big, but observers see the shape as representing four times as much. Not many people have tried to unify the field of information visualization – cartographers tend to analyse cartography, chart people talk about charts, computer scientists talk about on-screen design. However, Conrad mentioned a number of writers including Jacques Bertin, Edward Tufte, Jan V White, Gene Zelazny, Doug Simonds, Clive Richards, Michael Twyman, B Tversky, Robert Horn (Visual Language), Alan MacEachran (How Maps Work), Colin Ware (Information Visualization), and Yuri Engelhardt (The Language of Graphics).
He discussed symbology, compositionality, and Jacques Bertin’s “retinal variables” and presented various examples to illustrate how visual elements are put together – as a form of grammar, and how vocabularies such as line weights and thicknesses in network diagrams are another way of adding extra meaning to graphics.
Mapping SoftwareMartin Dodge of the University of Manchester, whose work includes the Atlas of Cyberspace and the Codespace project, discussed the problems of trying to represent the Internet as a map. Online spaces are not like Euclidean space, much of the web is “dark”- inaccessible for reasons of security, copyright protection, or just not easy to index. It is also so huge that you can’t download it into something else in order to run algorithms on it or process it in order to make a smaller or compressed model. One issue is that people download millions of web pages and throw algorithms across the data without really knowing what they want to see as a result. It is easy to create visualizations based on metaphors that actual confuse more than they clarify. One way to represent the web might be with 3D “fly thru” maps, but these are very hard to display on a flat screen.
Maps are an important part of our cultural psyche, determining how we see the world, but it is very easy to create maps and graphs that look wonderful but fail to convey any information. There are software packages that produce lovely visualizations of data, but without thought and user testing they may be meaningless or misleading.
Martin offered examples of cadastral maps – treemaps, hierarchical tiling, space filling maps, and landscape maps – built using clustering, self-organizing algorithms. He pointed out that these visualizations are possible but the Web looks pretty similar to how it did in 1993, with pages being the basic way information is accessed. We have not yet found a “tube map” for the Web and he questioned whether “cyberspace” is a bit of a red herring, and that it makes no sense to think of the Web as “space” at all.
He suggested that it might be more interesting to make a map of the Web and the world that would make sense to machines and that when we have ubiquitous computing and devices controlling the real world everywhere, we will need maps that those machines can use. James Birdle and Bruce Sterling have written in this area and there is a new field of research – software studies – with people like Matthew Fuller and Lev Manovich investigating how software is affecting culture and society. There is a “code space” that needs to work otherwise real world objects being controlled by software will literally crash. He suggested that rather than mapping the Web, it might be interesting to map a specific device, such as someone’s smartphone, which carries data about their movements, contacts, conversations, etc. and that we may end up with software that maps itself as it goes along. This could potentially have huge ethical implications. Other maps that might be interesting would be how a phone moves through “Hertzian space” - the world of electromagnetic waves and signals.
Touch technologiesMarianne Lykke of the University of Aalborg described an experiment to investigate the use of touch technologies in taxonomy building. It was interesting how themes mentioned by Conrad were reinforced by her findings. For example, touch technologies encourage egalitarian collaboration – like the drawings in sand – as quieter or shyer members of the group find it easier to join in with moving objects on a screen than speaking out in a group discussion or workshop.
Marianne explained how she had used touch technologies to replace card sorting techniques as part of the creative and analysis phase of designing a new knowledge organization system. The technologies were useful in finding user-friendly concepts, understanding user preferences, and selecting preferred terms.
She used Virtual Understanding Environments and Notebook, which were lightweight, easy to use, and to play with and interact with in groups. They helped participants more quickly come to a shared understanding than using paper-based techniques, enabled previous versions to be saved and stored more easily than with paper, and enabled embodied conversation as the group could work together.
Limitations were that only a few terms could be managed at a time and the technologies at the moment are only suitable for a small number of users, as the users have to be able to crowd round the screen.
Data journalismLisa Evans explained Data Journalism at The Guardian. She provided lots of examples of data-driven stories. One of the most popular is a visualization of government spending. As The Guardian has produced this every year, the team can now run comparisons and show trends over time.
They also build interactive tools – such as a tool that allowed you to “play chancellor” and allocate spending to different government departments. Data analysis can be used to check claims made by companies as well as governments, which the public would be interested to know. For example, they discovered that despite claims by the drinks company Innocent that they would donate some of the purchase price of bottles sold to charity, they had not in fact made any donations.
A map of where people involved in the 2011 riots were from showed that they were overwhelmingly from areas of social deprivation, despite government insistence that poverty was not a causal factor in the rioting.
The data team use a lot of tools, such as Many Eyes and Google Maps, but are also looking for Open Source tools. Better, faster tools and techniques are needed, as data analysis takes time but news demands rapid responses to events. Data journalism is in its infancy and already is proving very popular, with the data blog being the second most popular page on The Guardian’s website.
Follow the data team on twitter: @datastore @objectgroup.
Visual presentation of environment and human health data and informationWill Stahl-Timmins of the University of Exeter demonstrated his visualizations of health and climate change information and how these have been used to help people understand complex data. However, there is a danger that categorizations of data can be misleading, and with a graphic it is very easy to give a false impression.
He showed an example of a chart that showed UK carbon emissions by industry sector, but the “transport” category did not include international flights, which could lead people to jump to the wrong conclusions about the environmental impact of air travel.
Sometimes there has to be some “designers’ licence” to make graphics clear. He demonstrated a graphic that combined data about how close people live to sea with how they report their health. The trees were not in proportion to the distances from the sea, because they were not drawn to scale so that they would be visually pleasing and make it easy for viewers to spot where the “land” was.
He then talked us through a study assessing how well people understood a complex report about climate change written in words compared with a graphic representing the data in the report. People absorbed information form the graphic far faster than from the text, but it was not clear whether they had understood as much detail.